可使用的 ESRGAN 超分模型

最新推荐文章于 2025-07-23 11:31:23 发布

Covirtue

最新推荐文章于 2025-07-23 11:31:23 发布

阅读量1.9k

点赞数 1

CC 4.0 BY-SA版权

分类专栏： python 图片处理人工智能文章标签： python

本文链接：https://blog.csdn.net/weixin_47695827/article/details/138215208

Kaggle中使用

!pip install git+https://github.com/sberbank-ai/Real-ESRGAN.git

import os
from huggingface_hub import hf_hub_download
import torch
os.environ["HF_ENDPOINT"] = "https://hf-mirror.com"
model_path = hf_hub_download(repo_id="Shandypur/ESRGAN-4x-UltraSharp", filename="4x-UltraSharp.pth", repo_type="model")
weights = torch.load(model_path)  

# pth -> pth转换
map_key = [
    ["conv_body", "sub.23"],
    ["body", "sub"],
    ["rdb", "RDB"],
    ["", "model.1."],
    ["conv_first", "model.0"],
    ["conv_up1", "model.3"],
    ["conv_up2", "model.6"],
    ["conv_hr", "model.8"],
    ["conv_last"